Image to Video
Take a single still image and turn it into a short clip — the subject, lighting and composition you already approved carry through, with motion grafted on top. Our strongest i2v models today are Wan 2.7 I2V for polished commercial work and the Wan 2.2 i2v family (including a LoRA-aware ultra-fast variant) for stylized brand-specific motion.
Seedance 2.5 Text to Video is ByteDance's next-generation video model, turning a written brief into a single continuous take of up to 30 seconds — with dialogue, effects and ambience generated in the same pass. It renders up to native 4K and, when you need it, follows up to 30 reference images plus video and audio references for characters, products and motion.
Wan 3.0 Prime Reference to Video is the premium tier of Alibaba's Wan 3.0 reference mode, aimed at final deliverables rather than drafts. It builds a video of up to 30 seconds, with sound, from your own images, clips and audio — and is the tier to reach for when a character, a product label or a brand mascot has to stay locked across a reference-heavy shot.
Wan 3.0 Reference to Video turns the material you already have — product shots, character photos, a sample clip, a voice or a jingle — into a new video of up to 30 seconds with sound generated in the same pass. It is the standard tier of Alibaba's unified Wan 3.0 model and takes up to ten images, five video clips and five audio clips per shot, at 480p, 720p or 1080p.
Wan 2.2 i2v Fast is the PrunaAI-optimised variant of Alibaba's Wan 2.2 A14B image-to-video — engineered for short turnaround and predictable per-video pricing. Drop in a first frame, optionally a last frame, write a short motion prompt, and ship a 480p or 720p clip for ads, social cutdowns or product loops.
Wan 2.7 Image-to-Video animates a still image (or continues an existing clip) into a 3-15 second video, with optional first-and-last-frame control and audio synchronisation. Output runs at 720p or 1080p.
Seedance 2.0 Pro is ByteDance's flagship text-to-video model — strong prompt adherence, natural motion, and optional multimodal references (images, videos, audio) to lock identity, style, and rhythm. This is the native ByteDance ModelArk endpoint; the WaveSpeed-routed siblings stay in the catalogue for users that already point at them.
Pruna Video is a fast, multi-modal video generator — text-to-video, image-to-video and audio-to-video from a single endpoint. A built-in **draft mode** cuts the cost to about a quarter for rapid iteration; flip it off when you're ready for the final delivery take.
Pruna Video Avatar is the fastest talking-avatar / lip-sync model on the market — drop in a portrait photo plus either an audio clip or a written script, and get back a video of the subject speaking with synchronised lip movement, in 30 voices and 10 languages.
Happy Horse 1.0 is Alibaba's text- and image-to-video model. Generate a clip from a prompt, or animate a single still image into a fluid 3–15 second video at 720p or 1080p, with five aspect ratios for vertical, landscape and square delivery.